[layers] Rename lstm to rnn to be more general

bob/learn/tensorflow/layers/LSTM.py

@@ -4,32 +4,43 @@ import tensorflow as tf
 
 from tensorflow.python.layers import base
 
-def lstm(inputs, n_hidden, name=None):
+# def lstm(inputs, n_hidden, name=None):
+#     """
+#     """
+#     return LSTM(n_hidden=n_hidden, name=name)(inputs)
+
+def rnn(inputs, n_hidden, cell_fn, cell_args, name=None):
     """
     """
-    return LSTM(n_hidden=n_hidden, name=name)(inputs)
+    return RNN(n_hidden=n_hidden,
+               cell_fn = cell_fn,
+               cell_args = cell_args,
+               name=name)(inputs)
 
 
-class LSTM(base.Layer):
+class RNN(base.Layer):
     """
+    Inspired from tensorlayer/tensorlayer/layers.py
     """
 
-    def __init__(self,
-                 n_hidden,
+    def __init__(self, n_hidden,
+                 cell_fn = tf.nn.rnn_cell.BasicLSTMCell,
+                 cell_args = { "forget_bias": 1.0, },
                  name=None,
                  **kwargs):
-        super(LSTM, self).__init__(name=name,
-                                   **kwargs)
+        """
+        """
+        super(RNN, self).__init__(name=name, **kwargs)
+
         self.n_hidden = n_hidden
-        self.lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(self.n_hidden,
-                                                      forget_bias=1.0)
+        self.cell = cell_fn(num_units = self.n_hidden, **kwargs)
 
     def call(self, inputs, training=False):
         """
         """
-
-        outputs, states = tf.nn.static_rnn(self.lstm_cell,
+        outputs, states = tf.nn.static_rnn(self.cell,
                                            inputs,
                                            dtype=tf.float32)
 
+        # Compare to tensorlayer, it is as if return_last = True
         return outputs[-1]

bob/learn/tensorflow/layers/__init__.py

 from .Layer import Layer
 from .Conv1D import Conv1D
 #from .Maxout import maxout
-from .LSTM import lstm
+from .RNN import rnn
 
 
 # gets sphinx autodoc done right - don't remove it
@@ -21,6 +21,6 @@ def __appropriate__(*args):
 __appropriate__(
     Layer,
     Conv1D,
-    LSTM
+    RNN
     )
 __all__ = [_ for _ in dir() if not _.startswith('_')]

bob/learn/tensorflow/test/test_lstm.py

@@ -119,5 +119,6 @@ def test_dnn_trainer():
     embedding = Embedding(train_data_shuffler("data", from_queue=False), graph)
     accuracy = test_network(embedding, test_data, test_labels)
     logger.info("Accuracy {}".format(accuracy))
+    assert accuracy > 0.8
 
 test_dnn_trainer()